Chapter 9: Cold Noodles in Daejeon
Jang Seokho ate naengmyeon like he coded—fast, precise, and without wasted motion.
The noodle restaurant was a narrow, fluorescent-lit shop near KAIST’s main gate, the kind of place where the menu had three items and the ajumma didn’t ask what you wanted because everyone ordered the same thing. The naengmyeon arrived in steel bowls the size of serving platters, the buckwheat noodles coiled in ice-cold broth with thin slices of beef, half a hard-boiled egg, and julienned cucumber arranged on top like a mandala.
Dojun had taken the KTX from Seoul Station that morning—an hour and a half to Daejeon, then a bus to KAIST. The train ticket had cost thirty-two thousand won, which was a meaningful chunk of his contest prize money, but Seokho had insisted: “You haven’t lived until you’ve had the naengmyeon at Auntie Bong’s. Trust me.”
He hadn’t been wrong.
“You cut the noodles,” Seokho observed, watching Dojun use scissors on the long strands. “Most people just eat them whole.”
“I choke on long noodles.”
“That’s a character flaw.”
“It’s a survival instinct.”
Seokho almost smiled—the same thin, analytical expression that passed for a smile in his emotional vocabulary. He slurped his noodles with aggressive efficiency, then set down his chopsticks and crossed his arms.
“I’ve been analyzing your contest solutions,” he said, with the casual directness of someone who considered social pleasantries a waste of bandwidth. “I got the replay data from the contest system.”
“They give out replay data?”
“They do if you ask politely. Which, in my case, means asking once and then emailing the contest chair’s thesis advisor.” He pulled a printout from his bag and unfolded it on the table between the naengmyeon bowls. “Your Problem C implementation. The binary indexed tree approach.”
Dojun looked at the printout. It was his own code, annotated in Seokho’s neat, angular handwriting. Red circles around specific lines. Notes in the margins.
“Your BIT implementation is non-standard,” Seokho said. “Most people initialize from the bottom up. You initialized top-down with lazy propagation. It’s faster for the specific constraint in the problem—the bounded difference between consecutive elements—but it’s not a technique I’ve seen in any competitive programming resource.”
“I must have picked it up somewhere.”
“Where? I’ve read every major competitive programming textbook in Korean, English, and Japanese. I’ve gone through the entire Codeforces archive. This specific BIT variant doesn’t appear anywhere.” He tapped the printout. “So either you invented it independently, which would be remarkable, or you learned it from a source I haven’t found, which would be the first time that’s happened.”
Dojun ate his naengmyeon and considered his response carefully. The truth was that the BIT variant was something his team at Prometheus Labs had developed in 2024 for a specific distributed computing problem. He had used it in the contest without thinking, the way a musician might unconsciously play a riff they invented decades ago.
“I think I derived it,” he said. “When I was working through the problem, I realized the standard bottom-up approach would be slower for bounded differences, so I flipped the initialization order. It just… made sense.”
“It ‘just made sense,'” Seokho repeated. “Under contest pressure. In real time. You derived a non-trivial algorithmic optimization that I’ve never seen published, and it ‘just made sense.'”
“Yes.”
Seokho stared at him for a long moment. Then he folded the printout and put it away. “Okay, Park. I’m going to accept that answer. Not because I believe it, but because the alternative explanations are even less plausible.”
“What are the alternative explanations?”
“That you’re a time traveler from the future who already knows every algorithm that will ever be invented.” He said it deadpan, without a flicker of humor.
Dojun’s chopsticks froze halfway to his mouth.
“I’m joking,” Seokho added. “Obviously.”
“Obviously.”
“Although, for the record, if you were a time traveler, your strategy of pretending to be a mediocre sophomore while occasionally leaking genius-level insights would be the worst disguise in the history of time travel.”
“Good thing I’m not one, then.”
“Good thing.” Seokho picked up his chopsticks again. “Let’s talk about something else. What are you working on? Besides the contest circuit and whatever Kim Taesik has you doing.”
Dojun welcomed the subject change with visible relief. “A group project—campus navigation system with contextual pathfinding. A-star with modified heuristics that account for weather, crowd density, building schedules.”
“Contextual pathfinding.” Seokho’s eyebrows rose a fraction. “That’s ambitious for an undergraduate project. Whose idea?”
“My project partner’s. Lee Hana. She’s in the design track.”
“A designer came up with a contextual pathfinding concept?”
“She didn’t call it that. She said ‘best path instead of shortest path.’ Same idea, better language.”
“Hm.” Seokho considered this. “Better language is half the battle in product development. The other half is execution.” He paused. “What’s she like? This designer?”
“Brilliant. Perceptive. Direct to the point of being blunt. She reads people the way we read code—pattern matching, anomaly detection.” Dojun realized he was smiling and consciously flattened his expression. “She’s a good partner.”
“You’re smiling.”
“I’m not.”
“You were. Your facial muscles contracted in a configuration consistent with involuntary positive affect. I took a psychology elective.” Seokho’s deadpan cracked just slightly. “You like her.”
“We’re project partners.”
“Those two things aren’t mutually exclusive.” He ate the last of his egg. “I’ll stop. It’s not my business. But Park—a word of advice from someone who has exactly zero experience with romantic relationships and is therefore perfectly objective.”
“I’m listening.”
“Don’t let code be the only language you speak. I’ve watched brilliant people lose brilliant partners because they couldn’t say ‘I care about you’ without wrapping it in a function call.” He stood up and grabbed the bill before Dojun could reach it. “This is mine. You took the KTX. That’s enough financial damage for one day.”
“I can pay—”
“You can pay next time. In Seoul. I’ll find a place that’s worth the trip.”
After lunch, Seokho gave him a tour of KAIST. The campus was different from SNU—more modern, more compact, with a vaguely futuristic aesthetic that reminded Dojun of a tech company headquarters rather than a university. The buildings were connected by covered walkways and the paths were laid out with mathematical precision.
“This is the CS building,” Seokho said, gesturing at a glass-and-steel structure that looked like it had been designed by someone who really liked rectangles. “Third floor is the algorithms lab. That’s where I spend most of my time.”
“Can I see it?”
Seokho led him through a keycard-locked door and up a stairwell. The algorithms lab was a large open room with twelve workstations, whiteboards covering every wall, and the faint smell of coffee and desperation that was universal to academic computing spaces worldwide.
Three students were working, hunched over terminals. They looked up when Seokho entered, and their expressions shifted when they noticed Dojun.
“This is Park Dojun,” Seokho said. “SNU. The one from the contest.”
The reaction was immediate and gratifying—or it would have been, if Dojun wanted attention.
“You’re the guy who solved all five?” A student with round glasses and a Red Bull habit leaned forward. “Bro, your Problem E solution was insane. We’ve been trying to figure out how your pruning conditions worked. Dongwook thinks you used constraint propagation, but I think it’s some kind of—”
“It’s simpler than you think,” Dojun said. “The key was identifying which subtrees couldn’t possibly lead to valid solutions based on the register count constraint. Once you establish an upper bound on the remaining registers, you can prune any branch that exceeds it.”
“But how did you calculate the upper bound so efficiently? A naive bound would be too loose to prune anything useful.”
“I used the interference graph structure. If two code blocks interfere—meaning they need registers simultaneously—you can derive a tighter bound from the chromatic number estimate of the subgraph.”
Silence. The three KAIST students exchanged looks.
“That’s… chromatic number estimation,” Dongwook said slowly. “You used graph coloring theory inside a brute-force search.”
“A simplified version. I didn’t compute the actual chromatic number—that’s NP-hard. I used the degree-based upper bound. It’s fast enough for contest constraints.”
“You’re telling me you independently derived a hybrid approach that combines brute force with graph coloring heuristics, under contest time pressure, as a sophomore?”
“I had a good day.”
Seokho, who had been watching this exchange with his arms crossed and an expression that Dojun could only describe as smugly vindicated, said: “See? I told you he was interesting.”
“Interesting is an understatement,” the Red Bull student muttered. “Hyung, where did SNU find this guy?”
“That,” Seokho said, “is the question everyone keeps asking. And nobody, including Park himself, seems to have a satisfying answer.”
They walked around campus for another hour, talking about algorithms, about research, about the state of Korean computer science in 2006. Seokho was surprisingly easy to talk to when the competitive edge was set aside—curious, well-read, with opinions about everything from programming language design to the future of the internet.
“Web 2.0 is a bubble,” Seokho declared as they sat on a bench near the KAIST library. “All these social networking sites—Cyworld, MySpace—they’re building on advertising revenue. The moment the ad market contracts, they’ll collapse.”
“Some will survive,” Dojun said carefully. “The ones that build real network effects. Where the product becomes more valuable the more people use it.”
“Name one that will survive.”
“There’s a site called Facebook. It started at Harvard two years ago. It’s opening to the public later this year.”
“Facebook.” Seokho rolled the name around. “Never heard of it. What’s different about it?”
“Real identity. You use your actual name, your actual photo. It maps real-world social connections instead of creating anonymous ones. And it’s built on a platform model—other developers can build applications on top of it.”
“A platform model for social networking.” Seokho considered this. “That’s interesting. But it only works if the network reaches critical mass. How many users do they have?”
“About seven million, mostly American college students.”
“Seven million is nothing. Cyworld has twenty million in Korea alone.”
“Cyworld is a closed system. It can’t scale beyond Korea. Facebook’s architecture is designed for global scale from the start.” Dojun caught himself again—he was giving analysis that relied on knowledge of Facebook’s future. “At least, that’s what their public documentation suggests.”
“You read Facebook’s public documentation?”
“I told you. I read a lot.”
“There it is.” Seokho shook his head. “The Park Dojun catchphrase. You should trademark it.” He stretched his legs out on the bench. “Okay, hypothetical. If you could build anything right now—any product, any company—what would it be?”
The question hit like a spotlight. Dojun knew exactly what he would build. He had already built it once. Prometheus Labs—an AI-powered platform that made technology invisible, that anticipated human needs, that bridged the gap between what machines could do and what people actually wanted.
But he couldn’t say that. Not yet. Not to Seokho, who would remember every word and analyze it later.
“Something that makes technology easier for regular people,” he said. “Not for engineers or early adopters. For my mother. For people who don’t know what an algorithm is and don’t care, but whose lives could be better if the technology just… worked.”
“Technology for your mother.” Seokho raised an eyebrow. “That’s either very noble or very naive.”
“Why can’t it be both?”
“Because noble and naive usually end up bankrupt.” But his tone wasn’t dismissive—it was testing. Probing. “What would you build for your mother specifically?”
“An inventory system for her banchan stall. She tracks everything in her head—what sold, what’s running low, what to prepare tomorrow. She’s been doing it for thirty years and she’s never wrong, but she’s also exhausted. If I could build something that helped her manage the stall without having to learn anything new—no apps, no interfaces, just a system that fit into how she already works—that would be worth more than any social network.”
Seokho was quiet for a moment. Then: “That’s not naive. That’s actually a very difficult problem. You’re talking about ambient computing—technology that operates in the background of existing workflows without requiring behavior change.”
“Exactly.”
“In 2006, the hardware doesn’t exist for that. You’d need ubiquitous sensing, natural language processing, cloud infrastructure—none of which is ready.”
“Not yet. But it will be.”
“When?”
“Sooner than people think.”
Seokho studied him with that sharp, analytical gaze. “You keep saying things like that. ‘Sooner than people think.’ ‘It’s coming.’ Like you’ve seen a preview of the future that nobody else has access to.”
“Maybe I’m just optimistic.”
“You’re not. Optimists wave their hands. You give specifics—real identity social networks, platform models, ambient computing. These aren’t vague predictions. They’re roadmaps.” He stood up from the bench. “I don’t know what your deal is, Park. But I’m going to figure it out. That’s my new side project.”
“You have actual research projects.”
“This is more interesting.” He extended a hand. “Same time next month? You can show me Seoul.”
Dojun shook it. “Deal. But you’re buying the naengmyeon next time.”
“I paid today.”
“Seoul prices are higher.”
“Touché.” Seokho’s near-smile appeared again. “Take care, Park. And stop deleting your practice solutions. Some of us want to study them.”
On the KTX back to Seoul, Dojun pulled out his notebook and wrote.
Seokho observations:
1. He’s already suspicious. The BIT variant, the contest solutions, my Facebook analysis—each individually explainable, but collectively they form a pattern he’s tracking. I need to be more careful, or I need to commit to being less careful. The middle ground is shrinking.
2. He’s not an enemy. The original timeline’s rivalry happened because we were both building competing companies in the same space. Here, we’re just students. The competitive instinct is there, but it’s directed at problems, not at each other. This could be a friendship, if I don’t break it.
3. He’s lonely. He didn’t say it and would never say it, but the way he talked about his lab, his research, his routine—it was the portrait of someone who has colleagues but not friends. He came to Seoul to find me because he wanted a conversation with someone who operates on his level. That’s not rivalry. That’s a search for connection.
4. “Technology for your mother.” Why did I say that? Because it’s true. Because the products worth building are the ones that help people like Mom—people who don’t care about technology but whose lives are shaped by it. Prometheus Labs forgot this. I won’t.
He closed the notebook as the train pulled into Seoul Station. The evening was cool, the sky turning from blue to purple over the city’s concrete skyline.
His phone buzzed. A text from Hana:
How was Daejeon? Did the KAIST genius live up to the hype?
He typed back: He exceeded it. But he eats naengmyeon too fast. I don’t trust people who don’t savor their food.
LOL. Sounds like a CS student. All optimization, no appreciation. Hey – I started the wireframes for our portfolio project. Want to see them tomorrow?
Tomorrow works. Same jjigae place?
You just want more of Auntie’s jjigae.
Guilty.
Fine. Noon. Don’t be three minutes late this time.
He pocketed the phone and walked out of Seoul Station into the March night. The city was alive around him—neon, traffic, the smell of street food vendors setting up for the evening rush. A busker was playing guitar near the station exit, badly, and a small crowd had gathered more out of sympathy than appreciation.
Dojun dropped a thousand-won bill into the guitar case and kept walking.
Two relationships taking shape. Seokho, the rival becoming an ally. Hana, the partner becoming something he was afraid to name. Both of them seeing pieces of him that he hadn’t meant to show, both of them asking questions he couldn’t fully answer.
The original timeline was already fracturing. In his first life, he hadn’t met Seokho until his thirties. He hadn’t truly partnered with Hana until his late twenties. Now, at twenty, both connections were forming faster, deeper, and more honestly than before—and the divergence terrified him as much as it thrilled him.
Because every change meant the future he knew was becoming the future he didn’t. And a man who couldn’t predict what came next was just a man—no different from anyone else walking through Seoul Station on a Saturday night, hoping the path ahead was a good one.
The cherry blossoms were falling. Spring was almost over.
Whatever came next, he would face it without a script.